AI visibility for restaurants: named for tonight, with the right hours

What being named means when the question is about tonight

A guest's question has four parts: what they want to eat, a constraint (gluten-free, quiet, kids, a table for six), a time, and a place. An answer that names you has matched all four, and an answer that names you with the wrong closing time has sent a party to a locked door. So a restaurant's reading records the usual verdicts, named, recommended and first, and two more: what the answer named you for, and whether the facts it gave were right. Our methodology scores only the first of those verdicts, being named, and reports first place beside it.

Time is the part other trades do not have in the same way. A restaurant can be named for Friday dinner and left out of Sunday lunch, or named for tonight with last season's hours. The restaurants overview lists the facts a site has to state; this article is about finding out whether the answers use them.

Building a restaurant's panel

Write the time into the question as a day and an hour rather than "tonight" or "now", and the place as the name locals use rather than "near me", so the question means the same whenever and wherever it is asked. Ten to twelve questions is a full panel; setting up prompt monitoring gives the record format, and how to measure AI visibility covers the runs and the variation between them.

Checking the facts each answer gives

Read every answer twice: once for whether you were named, once for what it said. For a restaurant, the facts that go wrong are the ones that change: opening hours, the kitchen's last orders, holiday closures, whether you still serve brunch, a dish that has left the menu. Mark each fact right, wrong or out of date, and where the answer shows its sources, write down the page it came from. A wrong closing time comes from somewhere, on an old page, a profile or a reservation platform, and keeping a restaurant's data the same everywhere covers tracing it.

Menu questions need the same care. If an answer names you for a gluten-free pasta you no longer make, or misses the vegan main you added, the menu page is the first place to look; making your menu readable covers what the page needs.

Reading a first baseline

In the probe that accompanied our audit of 9,483 small business websites, one buyer question per business went to Claude with web retrieval on, and 14.9% of 9,240 businesses across 40 trades were named. That is one question, one assistant and one day, and not a restaurant figure. A restaurant's first reading that names it for few of its questions is the ordinary start.

Read it by kind of question. Named for your cuisine but never for a diet points at the menu's labels. Named on weekdays but never for Sunday points at the hours. Named when asked by name but not for your neighborhood points at the address and the pages that describe where you are. Each pattern is a page to fix, and the next month's reading shows whether the answers followed.

What the reading cannot count

Covers. A guest who heard your name from an assistant walks in or books through a platform, and nothing in the answer tells you so. Ask instead: a line on the booking form or a question at the host stand about how the guest found you, with an AI assistant among the answers, read beside the panel. And a reading covers only the questions in the panel, so build it from what your hosts are asked on the phone.